"Astra, a new model that OpenAI is testing internally, is amazing. No denying that:"But at the same time, a whole raft of people, some fairly prominent, are running around making a deeply flawed argument about the implications. ...
"[One] tweet went so far as to claim that “the species just crossed a one-way threshold”; Elon Musk took it as evidence that we had reached The Singularity.
"Each is making essentially the same error.
"The fallacy has a name; it’s called the fallacy of composition. ...
"What’s the manifestation of the fallacy in the current case? Thinking that a system that is great at a certain kind of maths problem is great at all maths, great at science or even quite possibly great at everything. The 'AGI-is-near' community keeps committing the same logical fallacy over and over. Every time there’s an advance, I see the same error.
"Here’s how the fallacy works."You don’t have to be a cognitive psychologist to realize that this inference just doesn’t follow."
- Someone pretends that all cognition is created equally. (Totally untrue.)
- Whenever AI achieves success on some form of (fancy) cognition, they want you to believe that success on all forms of AI is imminent.
~ Gary Marcus from his post 'OpenAI’s amazing — but vastly oversold — new model Astra'
▼
Monday, 3 August 2026
"The 'AGI-is-near' community keeps committing the same logical fallacy over and over. Every time there’s an advance, I see the same error."
1 comment:
We welcome thoughtful disagreement.
But we do (ir)regularly moderate comments -- and we *will* delete any with insulting or abusive language. Or if they're just inane or off topic. It’s okay to disagree, but pretend you’re having a drink in the living room with the person you’re disagreeing with. This includes me.
PS: Have the honesty and courage to use your real name. That gives added weight to any opinion.
But are these new proofs correct? I have not seen any confirmation of this.
ReplyDeleteIt was famous that Fermat's Last Theorem had many "proofs" that turned out to be wrong, before a correct (and enormously complicated) proof was found.
I will believe that this AI model has proved these conjectures when each proof has been published and has survived examination by suitably expert mathematicians. Until then, all we have is a PR claim.